#humanitariandata

[andrejverity]

Retracing Our Digital Footprints: my Humanitarian IM History Project

It is official! I am picking up a research project that I originally started in early 2023: milestones and stories from the 100-year evolution of humanitarian Information Management (IM). As we navigate rapid technological shifts and organizational changes, I believe there is significant value in looking back to better understand the trajectory of our discipline.

My Scope: From the League of Nations to Today

The initial focus of this research begins with the founding of the International Relief Union (IRU), as part of the League of Nations, and traces the lineage through the various iterations of UN coordinating entities from UNDRO to DHA to the OCHA we know today.

While the core of my research centers on these primary organizations, the timeline will also encompass innovations from other humanitarian entities when those milestones have significantly influenced the broader humanitarian IM discipline. To maintain a clear focus, I am intentionally excluding fields like Peacekeeping, Human Rights, and Mine Action, as well as agency-specific milestones. As with any research endeavor, I expect this scope may change and evolve as more historical context is uncovered.

My Roadmap: From Matrix to Stories

My first planned output will be a matrix of milestones, similar in format to my 200 Years of Technology Milestones. This will provide a high-level view of the major events, tools, and shifts that have defined our field over the last century - from the first thematic maps of the League of Nations in 1927 to the retirement of the Humanitarian Exchange Language (HXL) in 2026.

Over time, I hope to flesh this out by sharing the narratives and "stories" behind specific milestones. History is rarely just a list of dates; it is the story of people solving problems, like UNDRO staff in 1985 using "Le TUBE" pneumatic systems to rush urgent messages around the office, or what many consider the first use of a "Humanitarian Information Center" during the Kosovo response in 1999.

An Innovation Lens

My larger objective is to look at these historical markers through a lens of innovation. I would like to move beyond a simple chronology and collection of stories to ask the deeper questions:

What specific conditions or technologies enabled these milestones to happen?

Why could they not happen earlier?

How did these shifts fundamentally impact disaster relief efforts on the ground?.

Ultimately, I hope this research helps current and future IM professionals retrace the origins of the tools and practices they use every day. By understanding our shared history, we can better appreciate the 100 years of progress that have built our professional community.

Stay tuned as I begin sharing the first versions of the matrix and the stories behind them.

Andrej

[andrejverity]

Optimizing Data for AI Chatbots

We all know that AI is rapidly changing how people interact with information. And, whether we like it or not, AI chatbots will likely become the new “first reader” of the data we publish. For the humanitarian sector it means that:

Donors will increasingly use AI to check funding patterns.

Researchers will feed datasets into chatbots to explore trends.

Journalists will run quick analyses to verify claims.

Students and volunteers will  rely on AI to understand humanitarian operations.

And inside organizations, staff will use these tools to speed up analysis and decision‑making.

This shift creates a change in who interacts with our data. In the past, data was generally handled by experts - specialists who knew the quirks of a specific dataset and could manually correct for "bad data" as they worked. Today, we are seeing the rise of the "Initiator", a concept explained to me by Jerri Husch as non-data experts who use AI to jump straight into exploration. Since they do not have years of experience with the underlying spreadsheets, they might not realize when a merged cell or a cryptic abbreviation has led the AI to a false conclusion. 

So what’s the issue with AI reading humanitarian data? Because most publicly available humanitarian datasets, however well‑intentioned, are not structured in ways that AI systems can reliably interpret. Research showed that even the most advanced AI models can experience up to a 59% drop in accuracy when confronted with real‑world humanitarian datasets that contain common issues: merged cells, missing values, inconsistent formatting, unclear column names, extra rows, or oversized files. This means that AI-generated insights can easily become flawed: incorrect totals, missing rows, invented numbers, wrong trends, or misleading conclusions. Bad data can no longer be overlooked; it must be fixed at the source.

Thus, the question becomes: What can a data provider do to help increase the chances that AI interprets their data correctly? Together with Ralfs Rudzitis, we set out to explore this question with an overall objective of developing practical recommendations for those who provide downloadable* data. 

The good news is that the solutions are practical, accessible, and within reach for organizations of any size. They are:

Keep your data clean and simple. Delete unnecessary columns and rows. Avoid merged cells. Ensure column names are in the first row. Use one table per sheet.

Use clear, human‑readable column names. Use full words (e.g., Donor_Organization, not DonOrg). Avoid cryptic abbreviations or codes unless documented. Use underscores or spaces, not dots.

Provide context and metadata. If you provide an Excel file, consider adding a second sheet that: explains every column, clarifies abbreviations, provides basic guidance on how the dataset should be read.

Favor “wide” rather than “long” datasets. AI chatbots are surprisingly better with more columns and fewer rows than the opposite. If possible, structure or aggregate data accordingly to increase accuracy.

Include instructions in the dataset or accompanying page. Even a short paragraph can prevent major misinterpretations. What the dataset represents. What limitations exist. What filters or assumptions should (or should not) be applied. Which values indicate missing or invalid data.

For more details and a self-assessment checklist, see the full report Optimizing Data for AI Chatbots 2026

Why this matters for the humanitarian sector

Humanitarian data is a public good. We publish it so that people can use it to make better decisions, generate insights, and strengthen accountability. And now, non-experts are increasingly augmented with AI chatbots. If AI is going to analyze our data, and it will, then we must prepare our data for AI. This is not a luxury, but a modern data responsibility.

The good news is that it does not require new budgets, new software or new teams. Just better habits.

Andrej * While our findings focused on downloadable data and not (well structured) APIs or MCP servers, it is highly probable that many of the recommendations may still apply.

[canayatalay]

From insecurity of feeling naked on #iot to saving lives using #humanitariandata on #hdx?... How do we facilitate giving up control to masses, power to people? #entrepreneurship #startup #innovation #inspiration #iamcanay (at JW Marriott Austin)